| ← back to comparethecloud.net | | Compare the Cloud · Morning Edition | Thursday, 14 May 2026 · London |
Morning Edition.Ten curated stories, worldwide perspectives, before 9 a.m. | | 01 — Anthropic finally beat OpenAI in business AI adoption — but three big threats... | 01 | | 02 — Anthropic reinstates OpenClaw and third-party agent usage on Claude subscript... | 02 | | 03 — OpenAI publishes the sandboxing architecture that makes Codex safe to run on ... | 03 | | 04 — Google Chrome quietly downloaded 4 GB of AI model files to user machines with... | 04 | | 05 — A Windows 11 BitLocker bypass using only a USB stick has been published — and... | 05 | | 06 — AI IQ launches a public scorecard rating 50-plus frontier language models on ... | 06 | | 07 — Googlebook's confirmed Intel and Qualcomm partnership puts Microsoft... | 07 | | 08 — The EU is committing over €1 million to KDE as France's government quits... | 08 | | 09 — A hidden Windows Secure Boot certificate is about to expire — enterprise IT t... | 09 | | 10 — AMD brings 3D V-Cache to its Ryzen Pro 9000 workstation CPU series, unlocking... | 10 |
| | Worldwide · Enterprise AI | 01 |
CTC Newsroom Anthropic finally beat OpenAI in business AI adoption — but three big threats could erase its lead.For the first time, Anthropic's Claude has reached 34.4% adoption among businesses paying for AI tools, surpassing OpenAI's ChatGPT at 32.3%, according to the May 2026 Ramp AI Index — a crossover that marks a measurable shift in enterprise AI procurement. 34.4% Claude enterprise adoption | 32.3% OpenAI adoption | 50.6% Overall AI adoption |
| This crossover is strategically significant — it marks the moment Anthropic's enterprise focus translated into measurable procurement share. The three threats flagged — OpenAI model improvements, Google Workspace integration, and cost pressure — are real constraints, but the fact that half of all businesses now pay for AI tooling signals the end of the pilot era. For IT leaders, the operative question is no longer whether to standardise on an AI provider, but which one — and Claude's slim lead shifts that conversation considerably. — Kate Bennett · CEO, Compare the Cloud |
| | Worldwide · AI Governance | 02 |
Anthropic reinstates OpenClaw and third-party agent usage on Claude subscriptions — with a catch.Anthropic has reversed its brief ban on third-party agent harnesses such as OpenClaw, restoring access for paid Claude subscribers — but introducing a new compute-tiered subcategory that caps how much agent compute personal-plan users can draw. | The reinstatement is welcome news for engineering teams that built workflows on OpenClaw and similar open-source harnesses — the week-long access loss exposed how fragile subscription-policy decisions can be for enterprise teams with no SLA. The catch, however, is the new compute cap: organisations running high-frequency agentic workflows will need to reassess their subscription tier and budget accordingly. Watch for the Max tier to become the de facto standard for any team running more than a few hundred agent calls per day. — Kate Bennett · CEO, Compare the Cloud |
| | Global · AI Development Security | 03 |
· · · OpenAI publishes the sandboxing architecture that makes Codex safe to run on Windows.OpenAI has published a detailed technical overview of the sandboxing architecture built for Codex on Windows, describing how the system enforces controlled file access and network isolation to prevent the coding agent from executing unintended actions on developer workstations. | The publication of sandboxing architecture is a signal that enterprise AI coding tools are maturing beyond 'try it and see' into auditable, policy-compliant deployments. For security teams evaluating whether to permit AI coding agents on developer workstations, OpenAI's documented controls — file-system scope limits, network-exit restrictions, isolated process trees — are precisely the specificity that procurement and compliance reviews require. The Windows focus matters too: most enterprise developer fleets still run Windows, and until now sandboxed AI coding was largely a Linux and macOS conversation. — Kate Bennett · CEO, Compare the Cloud |
| | Worldwide · Enterprise MDM | 04 |
4 GB Google Chrome quietly downloaded 4 GB of AI model files to user machines without consent.0 User notifications sent | Windows + macOS Platforms affected |
| For enterprise IT teams managing standardised Chrome deployments, an undisclosed 4 GB download is a policy violation regardless of intent. Storage quotas on managed endpoints exist for compliance and operational reasons; software that silently consumes them creates audit gaps and support overhead. The deeper concern is precedent: if browser vendors treat on-device AI model installation as routine update behaviour rather than a material change requiring disclosure, MDM policies need updating now to explicitly govern browser-side AI asset downloads — before the next silent install is 40 GB rather than 4. — Kate Bennett · CEO, Compare the Cloud |
| Alert A Windows 11 BitLocker bypass using only a USB stick has been published — and the researcher suspects a backdoor.A security researcher has published a proof-of-concept demonstrating that Windows 11's BitLocker encryption can be bypassed using only a USB stick, with no specialist hardware required — and has claimed the mechanism may be an intentional backdoor rather than an implementation flaw. | If confirmed as a deliberate access mechanism rather than an implementation bug, this is a category-one incident for enterprise device policy. BitLocker is the primary at-rest encryption layer on the Windows estate; a reproducible USB-based bypass collapses the threat model that underpins laptop-loss policies, travel-security reviews, and GDPR-compliant device disposal. IT teams should treat this as unverified but material until Microsoft responds formally — accelerating BitLocker PIN policies and USB boot restrictions is prudent housekeeping regardless of the root cause. — Kate Bennett · CEO, Compare the Cloud |
| | Worldwide · AI Evaluation | 06 |
$ worldwide/ai evaluation AI IQ launches a public scorecard rating 50-plus frontier language models on a human intelligence scale.A startup project called AI IQ has published estimated intelligence quotients for more than 50 frontier language models, mapping them on a standard bell curve and making the interactive results publicly available for comparison across providers and versions. | The appeal of IQ-as-benchmark is obvious — a single number that cuts through the proliferation of leaderboards and task-specific evals. The problem is that intelligence quotients were contested measurement tools even when applied to humans; applying them to systems with fundamentally different failure modes risks anchoring procurement conversations to a metric that doesn't reflect actual deployment performance. That said, a public, comparable dataset of cross-model scores has genuine utility as a starting point — provided buyers treat it as one signal among several rather than a procurement shortcut. — Kate Bennett · CEO, Compare the Cloud |
| | Worldwide · Computing Platforms | 07 |
Googlebook's confirmed Intel and Qualcomm partnership puts Microsoft's PC platform position under serious pressure.Google's Googlebook laptop line has been confirmed to ship with both Intel and Qualcomm processors, with Intel describing the devices as 'premium, powerful devices designed for intelligence' — a positioning statement that directly challenges both Microsoft's Windows PC ecosystem and Apple's Mac lineup. Intel + Qualcomm confirmed chip partners |
| The Googlebook announcement is the clearest signal yet that Google intends to compete seriously in the premium enterprise laptop market rather than cede it to Apple and Microsoft. The dual-chip strategy — both Intel x86 and Qualcomm ARM — means Googlebook can serve the breadth of enterprise software compatibility requirements rather than a narrow ARM-native developer base. The strategic question for IT directors is whether Chrome OS has matured sufficiently to serve as a primary work OS for non-developer roles; Google's premium positioning suggests it is finally ready to make that argument. — Kate Bennett · CEO, Compare the Cloud |
| | Europe · Technology Policy | 08 |
Liberté, égalité, sovereignty. The EU is committing over €1 million to KDE as France's government quits Windows for Linux.The European Union has committed more than €1 million in grants to the KDE open-source desktop project as part of a broader drive to reduce public-sector dependence on proprietary software, a movement that has already seen France's government announce a switch from Windows to Linux on government workstations. | This is less about KDE and more about the EU's accelerating conviction that dependence on US software stacks is a strategic liability. The French decision to migrate government workstations to Linux is the canary; the KDE grant signals institutional money following political will. For European enterprise CIOs, the more immediate question is whether procurement-compliance pressure will eventually extend beyond the public sector — and whether the open-source desktop ecosystem is genuinely ready to absorb an institutional migration at scale. — Kate Bennett · CEO, Compare the Cloud |
| | Worldwide · Security Infrastructure | 09 |
| / / / / / / / / / / / / / / / / / / / / |
Zero-day A hidden Windows Secure Boot certificate is about to expire — enterprise IT teams need to audit affected devices now.Microsoft has issued warnings about a mandatory Secure Boot certificate renewal deadline arriving in 2026, with some Windows machines already displaying notices in the Device Security settings — failure to renew in time risks disrupting the boot chain on affected devices. | Secure Boot certificate expiry is the kind of issue that gets filed under 'someone else will fix it' right up until devices stop booting on a Monday morning. Enterprise IT teams managing heterogeneous Windows estates need to audit affected device models now — not after Patch Tuesday. The deadline is known, the fix is a firmware update, and the blast radius is entirely predictable: failure to act is an operational own goal on a calendar item. Check Device Security on representative samples today. — Kate Bennett · CEO, Compare the Cloud |
| | Worldwide · Hardware Procurement | 10 |
AMD brings 3D V-Cache to its Ryzen Pro 9000 workstation CPU series, unlocking higher performance for compute-intensive enterprise workloads.AMD has announced new Ryzen Pro 9000 series workstation CPUs with 3D V-Cache technology and higher TDP headroom, bringing the same cache-stacking architecture that dramatically improved workstation performance in the consumer Ryzen 9000 line to enterprise-grade, vPro-managed platforms. | Article I. Read the clause as you would a court ruling: the practical effect starts on publication, not the day the text was first circulated. |
| The arrival of 3D V-Cache on the Pro workstation line matters for IT procurement teams evaluating workstation refreshes — particularly for teams running simulation, CAD, video production, or scientific computing workloads that saturate conventional L3 cache. The vPro integration means these CPUs can slot into managed enterprise refresh cycles without a change in fleet management tooling. For procurement teams mid-cycle, the practical question is whether the performance uplift justifies accelerating the refresh cadence or whether it is a reason to delay purchases until Ryzen Pro 9000 availability stabilises. — Kate Bennett · CEO, Compare the Cloud |
| That's the front page.Curated from the CTC Monitor worldwide feed — narrowed to the ten that matter before nine. Morning Edition · Compare the Cloud · Thursday, 14 May 2026 · London View on the web · Unsubscribe |
|